Uncovering Critical Features for Deepfake Detection through the Lottery Ticket Hypothesis

2025-07-21

AI Summary

To address the weak interpretability, excessive parameter count, and deployment difficulty of deepfake detection models, this work pioneers the application of the Lottery Ticket Hypothesis to this domain. We propose an iterative magnitude pruning-based method to identify sparse “winning ticket” subnetworks and integrate Grad-CAM for interpretability validation. Evaluated on OpenForensics and FaceForensics++, our approach—using MesoNet as backbone—achieves 56.2% accuracy at 80% sparsity (retaining only 3,000 parameters), a mere 6.4 percentage points below the full-model baseline (62.6%), and significantly outperforms one-shot pruning. Crucially, the identified subnetworks demonstrate strong cross-dataset generalization. This work establishes a novel paradigm for lightweight, interpretable deepfake detection.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Learning on the Edge & Model Compression
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Recent advances in deepfake technology have created increasingly convincing synthetic media that poses significant challenges to information integrity and social trust. While current detection methods show promise, their underlying mechanisms remain poorly understood, and the large sizes of their models make them challenging to deploy in resource-limited environments. This study investigates the application of the Lottery Ticket Hypothesis (LTH) to deepfake detection, aiming to identify the key features crucial for recognizing deepfakes. We examine how neural networks can be efficiently pruned while maintaining high detection accuracy. Through extensive experiments with MesoNet, CNN-5, and ResNet-18 architectures on the OpenForensic and FaceForensics++ datasets, we find that deepfake detection networks contain winning tickets, i.e., subnetworks, that preserve performance even at substantial sparsity levels. Our results indicate that MesoNet retains 56.2% accuracy at 80% sparsity on the OpenForensic dataset, with only 3,000 parameters, which is about 90% of its baseline accuracy (62.6%). The results also show that our proposed LTH-based iterative magnitude pruning approach consistently outperforms one-shot pruning methods. Using Grad-CAM visualization, we analyze how pruned networks maintain their focus on critical facial regions for deepfake detection. Additionally, we demonstrate the transferability of winning tickets across datasets, suggesting potential for efficient, deployable deepfake detection systems.

Problem

• Identify key features for deepfake detection using Lottery Ticket Hypothesis
• Prune neural networks efficiently while maintaining high detection accuracy
• Transfer winning tickets across datasets for deployable detection systems

Innovation

• Applies Lottery Ticket Hypothesis to deepfake detection
• Identifies key features via iterative magnitude pruning
• Achieves high sparsity with maintained accuracy